Products in a retail store may be arranged at respective product positions along fixtures such as shelves, racks or other displays. The product positions may for example be selected to maximize sales by promoting certain products and/or to improve logistics of the store. Product positions for at least some products may be changed quite often and it may be difficult to maintain an updated record of the actual product positions.
Knowledge of the actual product positions may for example save time when products are restocked as persons bringing new items of the products to the fixtures of the store (e.g. from a storage room, a warehouse, or a supply truck) need not search through the fixtures of the store for the correct product positions at which to arrange the new items of the respective products. It is also important to detect low stock on the shelves as soon as possible since low stock on the shelves may lead to a lost sales opportunity for the retail store.
However, the resolution provided in images captured by cameras employed in many video-based security systems may be too low to detect the identity of a product that needs to be restocked. Therefore, unless the person performing the restocking brings items off all products that may potentially need to be restocked to the shelf, the person may need to go to the shelf to check which product to restock before bringing the appropriate items from a storage room, warehouse or supply truck. Knowledge of the actual product positions for the respective products may therefore make restocking more efficient.
Some objects may be difficult to detect or distinguish in images captured by a camera, for example due to low image resolution or poor illumination in an area (or room) in which the images were captured. Electronic labels, employed in retail stores for displaying information about products, may for example be difficult to distinguish in images captured by a camera. Although image processing sometimes allows objects to be detected in images, even when the image quality is rather poor, it would be advantageous to provide new methods for determining positions of electronic labels in images.